Jul 30, 2023
Comprendre non
Scientific Reports volume 13, Numéro d'article : 12524 (2023) Citer cet article 216 Accès 1 Détails de Altmetric Metrics Le changement climatique a un impact significatif sur le cycle hydrologique mondial, conduisant à
Rapports scientifiques volume 13, Numéro d'article : 12524 (2023) Citer cet article
216 accès
1 Altmétrique
Détails des métriques
Le changement climatique a un impact significatif sur le cycle hydrologique mondial, entraînant des changements prononcés dans les extrêmes hydroclimatiques, tels qu'une augmentation de la durée, de l'occurrence et de l'intensité. Malgré ces changements importants, notre compréhension des risques hydroclimatiques et de la résilience hydrologique reste limitée, en particulier à l’échelle des bassins versants de la péninsule indienne. Cette étude vise à combler cette lacune en examinant les extrêmes hydroclimatiques et la résilience dans 54 bassins versants péninsulaires de 1988 à 2011. Nous évaluons dans un premier temps les indices de précipitations et de débits extrêmes et estimons les niveaux de retour de conception à l'aide de modèles non stationnaires de valeurs extrêmes généralisées (GEV) qui utilisent le climat mondial. modes (ENSO, IOD et AMO) comme covariables. De plus, la résilience hydrologique est évaluée à l'aide d'un modèle convexe qui intègre le débit simulé du meilleur modèle hydrologique parmi SVM, RVM, forêt aléatoire et un modèle conceptuel (abcd). Notre analyse montre que les configurations spatiales des indices moyens de précipitations extrêmes (R1 et R5) ressemblent pour la plupart aux indices de débit extrêmes (Q1 et Q5). De plus, tous les indices extrêmes, notamment R1, Q1, R5 et Q5, démontrent un comportement non stationnaire, indiquant l'influence substantielle des modes climatiques mondiaux sur les précipitations et inondations extrêmes dans les bassins versants. Nos résultats indiquent que le modèle de forêt aléatoire surpasse les autres. Par ailleurs, nous constatons que 68,52% des bassins versants présentent une résilience hydrologique faible à modérée. Nos résultats soulignent l’importance de comprendre les risques hydroclimatiques et la résilience des bassins versants pour des prévisions précises de l’impact du changement climatique et des stratégies d’adaptation efficaces.
Les impacts projetés des extrêmes hydroclimatiques ont des implications en termes de risque et de résilience à l’échelle du bassin versant. En évaluant les extrêmes de précipitations et de débits, il est possible de mieux comprendre les effets potentiels du changement climatique, puisque les précipitations sont le principal processus hydrologique, tandis que le débit représente une réponse agrégée de différentes variables hydroclimatiques à l'échelle d'un bassin versant1,2. Face à de graves perturbations telles que les inondations et les sécheresses dues au changement climatique3,4, il est essentiel de comprendre la capacité de résilience des bassins versants, en particulier l'ampleur des perturbations auxquelles ils peuvent résister et dont ils peuvent se remettre. Par conséquent, les études à l’échelle des bassins versants sont cruciales pour évaluer les impacts du changement climatique sur la résilience des bassins versants et les extrêmes hydroclimatiques. Ces études peuvent jouer un rôle clé dans l’évaluation des risques et de la vulnérabilité des systèmes de ressources en eau.
De nombreuses études ont appliqué la théorie des valeurs extrêmes (EVT) pour évaluer les risques associés aux précipitations extrêmes ou aux inondations5,6. L'EVT implique un processus en deux étapes. Dans un premier temps, les événements extrêmes, tels que les inondations, sont isolés à l’aide de méthodes telles que les maxima de bloc ou la méthode Peaks Over Threshold. Par la suite, une distribution de probabilité théorique, telle que la valeur extrême généralisée ou la distribution de Pareto généralisée, est adaptée à cette série séparée d'événements extrêmes. À partir de cette distribution ajustée, l'ampleur du débit de conception requis est déterminée afin d'évaluer le risque associé. Traditionnellement, de nombreuses études ont adopté une approche stationnaire, dans laquelle les paramètres de la distribution ajustée à l'aide de l'EVT sont supposés rester constants dans le temps7. Cependant, compte tenu de l’évolution constante du climat et de ses effets dynamiques sur le système hydroclimatique, cette approche stationnaire pourrait ne plus suffire à l’évaluation des risques8. Les non-stationnarités introduites par les changements climatiques et anthropiques pourraient conduire à des erreurs substantielles dans l’estimation des niveaux de rendement extrêmes9. Ces inexactitudes peuvent donner lieu à d’importantes erreurs d’estimation, sous-estimant ou surestimant la probabilité d’événements extrêmes, déformant ainsi le risque associé. Par exemple, l’inondation de Boulder, au Colorado, en 2013, a dépassé les prévisions du modèle stationnaire, sous-estimant le risque d’inondation10. Cela indique la nécessité d’approches non stationnaires, équilibrant les prévisions de risque tout en reconnaissant les incertitudes inhérentes8,11. Pour combler cette lacune, une approche non stationnaire est intégrée à la modélisation des événements extrêmes à l'aide de l'EVT. Récemment, plusieurs études ont été menées en utilisant une approche non stationnaire pour comprendre le comportement des extrêmes12,13.
1300 mm) in 33.33% of catchments each. Moderate to high mean prcptot primarily characterizes central-eastern, southernmost Western Ghats, and parts of the western region. On the other hand, central and southeastern regions predominantly exhibit low prcptot values (Fig. 1). Trend analysis reveals a positive trend in 53.70% of catchments, with a magnitude of 0–5/> 5 mm/year in 24.07/29.63% catchments, predominantly lying in upper central-eastern, western ghats and southern region. Conversely, 46.30% of catchments show a negative trend, with 18.52/27.78% having a trend magnitude of < − 3/− 3–0 mm/year, mainly in central and lower central-eastern regions. Notably, the spatial distribution of mean qtot closely resembles that of mean prcptot. Mean qtot also shows low (< 200 mm), moderate (200–500 mm), and high (> 500 mm) in 33.33% catchments each, identical as mean prcptot. The central, central-eastern, southernmost western ghats and parts of western region mostly exhibit a moderate-high mean qtot, whereas parts of southern-eastern region indicate a lower mean qtot. Trend analysis shows a positive trend in 40.74% of catchments, with a rate of 0–5/ > 5 mm/year in 27.78/12.96% catchments, primarily in small catchments. On the other hand, 59.26% of catchments show a negative trend, with 33.33/25.93% catchments having a trend magnitude of < − 5/− 5–0 mm/year, dispersed across the study area. Central and central-eastern regions showcase a few catchments with a negative trend in both prcptot and qtot. Overall, the spatial patterns identified in the trend analysis of prcptot and qtot slightly resemble each other./p> 70 mm), mostly situated in the south-eastern and parts of the central region. The R1 trend is positive in 51.85% of catchments, with 20.37% showing a trend magnitude of 0–0.75 mm/year and 31.48% exceeding 0.75 mm/year, mostly in the central-eastern and southernmost regions. On the other hand, R1 shows a negative trend in 48.15% of catchments, with magnitudes of < − 0.5 mm/year for 31.48% and between − 0.5 to 0 mm/year for 16.67% catchments, largely in the central region. The spatial pattern of mean Q1 closely aligns with that of mean R1. Around 62.96% of catchments show moderate to high magnitude Q1 (42.59% between 20–60 mm and 20.37% exceeding 60 mm), primarily in central, central-eastern regions, parts of southern Western Ghats, and a few catchments in the western region. The remaining 37.04% of catchments have a low mean Q1 (< 20 mm), typically moderate-large sized catchments in central and south-eastern regions. The Q1 trend is positive in 40.74% of catchments (25.93% showing 0–0.3 mm/year and 14.81% exceeding 0.3 mm/year), primarily in central-eastern and a few southernmost catchments. The remaining 59.26% of catchments exhibit a negative trend, with magnitudes of < − 0.3 mm/year for 29.63% and between − 0.3 to 0 mm/year for 29.63%, mostly in the central region. Overall, the spatial distribution of R1 and Q1 trends closely resemble each other./p> 100 mm). Additionally, 40.74% of catchments show a positive trend (16.67% recording 0–0.5 mm/year and 24.07% exceeding 0.5 mm/year), mostly in small to moderate-sized catchments. The remaining 59.26% show a negative trend, with magnitudes of < − 0.5 mm/year for 37.04% and between − 0.5 to 0 mm/year for 22.22% of catchments, predominantly in the southernmost, central, and parts of upper centre-eastern regions. In general, R5 and Q5 trends exhibit high resemblance except in parts of central-eastern region./p> 160 mm) in 14.81%, 53.70% and 31.48% catchments, respectively. It indicates a similar spatial pattern as mean R1 (Fig. 2), showing a moderate-high magnitude in central, central-eastern and southern western ghats region. Additionally, 50yrR1and 100yrR1, each show high magnitude (> 250 and > 300 mm) in 29.63% catchments. As for Q1, the best models are M11 (25.93% of catchments), M7 (22.22%), and other NS models (51.85%). The spatial pattern of 10yrQ1 show high resemblance with 10yrR1, with 25.93%, 48.15% and 25.93% of catchments exhibiting low (< 30 mm), moderate (30–80 mm), and high (> 80 mm) magnitude, respectively. Similarly, 50yrQ1/100yrQ1 attain high magnitude (> 140 and > 200 mm) in 27.78% of the catchments each. Consequently, it can be inferred that the spatial patterns of mean R1, 10yrR1, mean Q1 and 10yrQ1 show resemblance with each other./p> 350 mm) categories in 29.63%, 48.15% and 22.22% of catchments, respectively. The spatial distribution of 10yrR5 is similar to that of mean R5, with the moderate-high 10yrR5, primarily observed in central, central-eastern, southernmost western ghats and parts of the western region, while low 10yrR5 is mostly found in parts of the south-eastern region. Moreover, 50yrR5 and 100yrR5 show high magnitude (> 530 and > 650 mm) in 22.22 and 25.93% of catchments respectively. For Q5, the best models identified are M7 (27.78% of catchments) and M12 (16.67%), while other NS models are best fit for the remaining catchments (55.56%).10yrQ5 show low (< 85 mm), moderate (85–200 mm), and high (> 200 mm) magnitude are 29.63%, 40.74% and 29.63% of catchments, respectively. The spatial distribution of 10yrQ5 shares a high resemblance with 10yrR5. It is also evident that the patterns exhibited by mean R5, 10yrR5, mean Q5 and 10yrQ5 illustrate resemblance among each other./p> 0.80), good (0.65–0.80), acceptable (0.5–0.65) and bad (< 0.50) categories based upon Nash Sutcliffe Efficiency (NSE). The NSE statistics has been used extensively for defining model performance39 and therefore enables us to compare our results with earlier studies. For abcd model of 54 catchments, the calibration is carried out in previous study40 and results are discussed here briefly. In abcd model, 81.5% of the catchments show more than bad performance limit (NSE > 0.50) in calibration period, however only 60% of catchments performed above this range (Fig. 3). For random forest model, all catchments performed above bad range in calibration as well in validation period. In case of SVM, 94.4% catchments show above bad performance in calibration period, whereas the number of catchments performing above bad range plummets to 81.48% in validation period. Similarly, RVM shows similar behaviour with 92.5% catchments in validation period and 88.88% catchments in calibration period performing above bad range. Furthermore, the overall performance is considered as minimum of NSE statistics in calibration and validation period and is categorized as excellent (> 0.80), good (0.65–0.80), acceptable (0.5–0.65) and bad (< 0.50) as shown in Fig. 3. Overall performance of abcd model indicates 57% catchments have acceptable (12), good (13) and excellent scores (6). Overall performance of SVM/RVM model are similar and indicates 81.48/88.88% catchments have acceptable (21/18), good (16/21) and excellent scores (7/9)./p> 2.5) based on their mean resilience. A majority (68.52%) of catchments, mainly in the central, central-eastern, western ghats, and southern regions, fall into the low (11.11%) and moderate (57.41%) resilience categories, while 31.48% of catchments demonstrate high resilience (Fig. 4). Subsequently, the resilience variability is assessed using the coefficient of variation, leading to a classification into low (< 0.75), moderate (0.75–1), and high (> 1) variability. Notably, 38.89% and 25.93% of catchments display moderate and high variability, respectively. Additionally, a strong inverse relationship is observed between mean resilience and resilience variability (ρ = − 0.90), suggesting that catchments with high mean resilience typically exhibit low variability and vice versa. Furthermore, trend analysis on the resilience index time series reveals that while 33.33% of catchments show a negative trend, the remaining 66.67% demonstrate a positive trend. And it is to be noted that 48.15% and 18.52% of catchments exhibit a moderate (0–0.10/year) and high trend magnitude (> 0.10/year), respectively./p> 900 mm and > 200 mm) of mean annual precipitation (prcptot) and mean annual discharge (Qtot), respectively. Furthermore, a positive trend is observed in prcptot/Qtot across 53.70%/40.74% of the catchments, with both indices sharing similar spatial trend patterns. For R1 and Q1, moderate to high mean values (> 70 mm and > 20 mm respectively) are evident in 70.37%/62.96% of the catchments, primarily in the central, central-eastern, southern western ghats, and some catchments in the western region. Positive trends for R1/Q1 are present in 51.85%/40.74% of the catchments, exhibiting high resemblance in their spatial trend patterns. Additionally, mean R5 and Q5 also show an identical spatial pattern where 74.07%/64.82% of catchments demonstrate moderate to high magnitudes (> 150 mm and > 50 mm), respectively, predominantly lying-in centre-eastern, southernmost western ghats and parts of western region. And R5/Q5 indicate positive trend in 53.70/40.74% catchments. In general, spatial pattern of trend statistics of R5 and Q5 exhibit high resemblance except in central-eastern region. Overall, the spatial patterns of mean R1, Q1, R5, Q5, and Qtot mirror those of the mean annual precipitation (mean prcptot) across the 54 catchments. This indicates that prcptot is the primary hydrological process influencing hydroclimatic extremes. Overall, the spatial patterns of mean R1, Q1, R5, Q5 and qtot show slight resemblance with mean annual average precipitation across the 54 catchments, which indicates that precipitation is main hydrological process driving hydroclimatic extremes as evident in recent literature43,44,45./p> 100 mm/> 30 mm). Notably, the spatial patterns of mean R1, mean Q1, 10yrR1 and 10yrQ1 show strong resemblances to one another. Similarly, 70.37% of catchments exhibit moderate to high 10yrR5 and 10yrQ5 (> 220 mm and > 85 mm), and the spatial patterns of mean R5, mean Q5, 10yrR5, and 10yrQ5 also correspond closely with one another./p>2.0.CO;2" data-track-action="article reference" href="https://doi.org/10.1175%2F1520-0493%281988%29116%3C2417%3ASSBOTM%3E2.0.CO%3B2" aria-label="Article reference 39" data-doi="10.1175/1520-0493(1988)1162.0.CO;2"Article ADS MathSciNet Google Scholar /p>
Français
English
Deutsch
Español
Italiano
Português
日本語
한국어
Русский
Maison